user-mode-linux-devel

[User-mode-linux-devel] Re: 2.3.51 Networking *smile*

[William S. <wst...@po...>]

Good evening, Jeff,

On Sun, 12 Mar 2000, Jeff Dike wrote:

> > ifconfig umn 172.20.0.244 pointopoint 172.20.0.253
> > 	From which the uml-kernel could extract the pointopoint address and
> > execute
> 
> > um_ifconfig sl0 172.20.0.253 pointopoint 172.20.0.244 
> 
> I'll look through the networking code again, but when I wrote the driver, my 
> impression was that the only way to give the driver an address was in the 
> hardware address, which is why I sneak the IP address in via the ethernet 
> address.

	LOL!  You're kidding - I just went back and did the conversion and
only now reconized that the mac address encoded the IP address.

> What I'm going to do about this is add a command line arg like
> 	umn0=172.20.0.253
> 
> That will let you specify both ends of the slip device (the other is in the 
> ifconfig command at boot time).

	How would it be if you used ethertap on the host and uml?  Could
the uml kernel connect the output of one tap to the input of the other and
vice-versa?  Granted, this is still a point-to-point approach, but A) I
think it's still pretty neat, and B) it would mean the umn device wouldn't
have to be used at all.
	Ethertap's part of any kernel that would support uml and it seems
to work pretty reasonably; Rusty uses it for his netfilter
testing.  Because you'd be feeding the output of one into the input of the
other, it almost seems as if you could setup pipes to do the work.

> > 	I suppose Nirvanha would be to have sl0 on the host be able to talk
> > to multiple UML's on a simulated ethernet cable: 
> > host OS sl0: 192.168.12.1/24 
> > uml1 umn:	192.168.12.2/24, default route ->1 
> > uml2 umn:	192.168.12.3/24, default route ->1 
> > uml3 umn:	192.168.12.4/24, default route ->1 
> 
> You're more of a network weenie than I, but I thought that slip devices are 
> ptp, so each one has two end, and only one machine at each end.  My plan for a 

	Ooops, I wasn't paying attention.  You're exactly right.  It would
be ethx on the host and in the umn's.

> virtual net has been to use the host machine as a virtual hub rather than as a 
> virtual ethernet.  There would be a slip device for each uml, and the host 
> machine would be routing packets between them.  And for a truly virtual net, 
> I'm planning on a second driver which just uses normal pipes for umls to talk 
> directly to each other without routing through the host.

	Again, would the ethertap serve this purpose?

> > How many fhd's are supported?  It looks like MAX_DEV is set to 8 in
> > the uml patch.
> 
> Is that not enough for you?

	It was a poorly worded real question, not request.  8 is fine,
especially if adjustable.  I just wanted to check that I was reading it
correctly.

> > Do they support partition tables so one might actually mount an entire
> > physical hard drive via fhd:
> > ./linux fhd1=/dev/hdc
> > 	where /dev/hdc is a partitioned ide drive?
> 
> What good would this be?  The host OS already splits the partitions up and 
> makes them available individually.

	As one possibility, one could assign an ndb device to fhd1.  The
fake device at the other end of the network could be (perhaps _must_ be; I
don't know) pointing to a physical device with a partition table.
	Although I'll side with you on limited utility, I still bring it
up with the goal of making uml as close to the real thing as possible.

> > BTW - what do you think of the term Virtual Block Device as opposed to
> > Fake Hard Disk.  The former makes me think that I could map vbda=/dev/
> > hdc vbdb=/dev/ndb1 /vbdc=/dev/sr0 /vbdd=/dev/fd1 - anything that is a
> > block device, whereas fhd sounds like it can only be a hard disk.
> 
> I like that.  I'll switch at some point.
> 
> > Unimplemented syscall : 8
> >                          Untested (20092) [0x10171428]: syscall_kern.c
> > line 696
> 
> > 	From unistd.h: #define __NR_creat                8 
> 
> OK, I'll toss that in as well.

	It sounds like putting these in place is not all that complex.  
(Didn't I hear once that the mark of the true artist is one that makes the
impossible look effortless?  I think that's why you rank as a True Hacker
in my book.)  If it really is a matter of cut and paste, would it be
useful to simply enable the remaining syscalls and printk/syslog their use
the first time they get called in a given kernel along with a request to
mail "syscall 674, at location blah..." to you/the devel list?

> > Then start filling up ram with bash's.  Around 312K free, the xterm
> > with the bashs and the xterm with top freeze, but the console still
> > accepts keystrokes:
> > usermode:~# 
> > usermode:~# ps axf | less 
> Kernel panic: No vma in segv
> 
> I run with swap all the time, but I haven't actually forced it to swap in a 
> while.  I'll check this out.

	I brought it up because I'm moderately confident that I've never
successfully gotten any data into swap space on 2.3.51/2.3.51-uml.  

> > Jeff - could this make its way into cvs, the downloads section (in
> > text and html format; I'd be happy to provide a tar with all three if
> > that's useful), and into the debian root_fs package on next release
> > (it's still at 0.0.4)?
> 
> I forgot to rebuild the text versions of the HOWTO when I redid the packages.  
> Next time... :-)  Your last changes are available in all their glory in the 
> html version on the web site...

	I hadn't looked there, sorry.
	I'm actually a little surprised that Sourceforge doesn't have a
documentation section - they have everything else, and even with cute
little icons.  :-)
	Cheers,
	- Bill

---------------------------------------------------------------------------
admin@loopback, $LOGIN@localhost, $LOGNAME@localhost,
$USER@localhost, $USER@$HOST,-h1024@localhost,ro...@ma...
Chairman Reed Hundt: rh...@fc...
Commissioner James Quello: jq...@fc...
Commissioner Susan Ness: sn...@fc...
Commissioner Rachelle Chong: rc...@fc...
US Postal Service: cus...@em...
(Courtesy of Arlen Fletcher <fle...@pa...>)
--------------------------------------------------------------------------
William Stearns (wst...@po...).  Mason, Buildkernel, named2hosts, 
and ipfwadm2ipchains are at:                http://www.pobox.com/~wstearns
LinuxMonth; articles for Linux Enthusiasts! http://www.linuxmonth.com
--------------------------------------------------------------------------

[User-mode-linux-devel] Re: 2.3.51 Networking *smile*

[Jeff D. <jd...@ka...>]

> ifconfig umn down
> 	successfully returns to the command prompt, but promptly hangs uml
> with a: umn_read had error, errno = 9
> 	on the console and the idle thread at 99% on the host. 

I've seen this on shutting down, but it didn't hang anything.  I'll have a 
closer look at it.

> ifconfig umn 172.20.0.244 pointopoint 172.20.0.253
> 	From which the uml-kernel could extract the pointopoint address and
> execute

> um_ifconfig sl0 172.20.0.253 pointopoint 172.20.0.244 

I'll look through the networking code again, but when I wrote the driver, my 
impression was that the only way to give the driver an address was in the 
hardware address, which is why I sneak the IP address in via the ethernet 
address.

What I'm going to do about this is add a command line arg like
	umn0=172.20.0.253

That will let you specify both ends of the slip device (the other is in the 
ifconfig command at boot time).

> 	I suppose Nirvanha would be to have sl0 on the host be able to talk
> to multiple UML's on a simulated ethernet cable: 
> host OS sl0: 192.168.12.1/24 
> uml1 umn:	192.168.12.2/24, default route ->1 
> uml2 umn:	192.168.12.3/24, default route ->1 
> uml3 umn:	192.168.12.4/24, default route ->1 

You're more of a network weenie than I, but I thought that slip devices are 
ptp, so each one has two end, and only one machine at each end.  My plan for a 
virtual net has been to use the host machine as a virtual hub rather than as a 
virtual ethernet.  There would be a slip device for each uml, and the host 
machine would be routing packets between them.  And for a truly virtual net, 
I'm planning on a second driver which just uses normal pipes for umls to talk 
directly to each other without routing through the host.

> How many fhd's are supported?  It looks like MAX_DEV is set to 8 in
> the uml patch.

Is that not enough for you?

> Do they support partition tables so one might actually mount an entire
> physical hard drive via fhd:
> ./linux fhd1=/dev/hdc
> 	where /dev/hdc is a partitioned ide drive?

What good would this be?  The host OS already splits the partitions up and 
makes them available individually.

> BTW - what do you think of the term Virtual Block Device as opposed to
> Fake Hard Disk.  The former makes me think that I could map vbda=/dev/
> hdc vbdb=/dev/ndb1 /vbdc=/dev/sr0 /vbdd=/dev/fd1 - anything that is a
> block device, whereas fhd sounds like it can only be a hard disk.

I like that.  I'll switch at some point.

> Unimplemented syscall : 8
>                          Untested (20092) [0x10171428]: syscall_kern.c
> line 696

> 	From unistd.h: #define __NR_creat                8 

OK, I'll toss that in as well.

> Then start filling up ram with bash's.  Around 312K free, the xterm
> with the bashs and the xterm with top freeze, but the console still
> accepts keystrokes:
> usermode:~# 
> usermode:~# ps axf | less 
Kernel panic: No vma in segv

I run with swap all the time, but I haven't actually forced it to swap in a 
while.  I'll check this out.

> Jeff - could this make its way into cvs, the downloads section (in
> text and html format; I'd be happy to provide a tar with all three if
> that's useful), and into the debian root_fs package on next release
> (it's still at 0.0.4)?

I forgot to rebuild the text versions of the HOWTO when I redid the packages.  
Next time... :-)  Your last changes are available in all their glory in the 
html version on the web site...

				Jeff

[User-mode-linux-devel] Howto additions -> 0.0.6

[William S. <wst...@po...>]

Good evening, Jeff and all,
	I've made some changes to the howto.
- Included myself in the blame for the content. :-)
- Updated current kernel version to 2.3.51
- Minor changes to "starting uml" section.
- "Fun tricks" section.
	- Allocating ram to uml
	- Using fhd1 as a swap device
	- Step by step guide to getting networking running, up to and
including full masqueraded Internet access.

	Feedback is, of course, welcome.  Jeff - could this make its way
into cvs, the downloads section (in text and html format; I'd be happy to
provide a tar with all three if that's useful), and into the debian
root_fs package on next release (it's still at 0.0.4)?
	Cheers,
	- Bill

---------------------------------------------------------------------------
	"There are two kind of people, those how work and those how want the
credits. Try to stay in the first category. The competition is much
smaller."
	-- Mahatma Ghandi
--------------------------------------------------------------------------
William Stearns (wst...@po...).  Mason, Buildkernel, named2hosts, 
and ipfwadm2ipchains are at:                http://www.pobox.com/~wstearns
LinuxMonth; articles for Linux Enthusiasts! http://www.linuxmonth.com
--------------------------------------------------------------------------

[User-mode-linux-devel] Actual use of swap?

[William S. <wst...@po...>]

Good afternoon, Jeff,
	It's not clear whether the swap is actually working or not in
2.3.51-uml:
dd if=/dev/zero of=uml-swap1 bs=1M count=4
mkswap -c uml-swap1
./linux-2.3.51 fhd1=uml-swap1
swapon /dev/disk/1

	Then start filling up ram with bash's.  Around 312K free, the
xterm with the bashs and the xterm with top freeze, but the console still
accepts keystrokes:

usermode:~#
usermode:~# ps axf | less
Kernel panic: No vma in segv

	The most recently executed bash, by the way, appears to be the one
chewing up 99% CPU.
	I sincerely hope these reports are repeatable as I really have few
skills in gdb.
	Cheers,
	- Bill

---------------------------------------------------------------------------
	Weinberg's Law: If builders built buildings the way programmers
wrote programs, then the first woodpecker that came along would destroy
civilization. 
(Courtesy of David E. Vandewalle, van...@pr...)
--------------------------------------------------------------------------
William Stearns (wst...@po...).  Mason, Buildkernel, named2hosts, 
and ipfwadm2ipchains are at:                http://www.pobox.com/~wstearns
LinuxMonth; articles for Linux Enthusiasts! http://www.linuxmonth.com
--------------------------------------------------------------------------

[User-mode-linux-devel] (no subject)

[William S. <wst...@po...>]

Good afternoon, Jeff,
	2.3.51-uml Sitting relatively idle after swapon /dev/disk/1 and
loading up a few bash's in xterm1.  Possibly mistranscribed as it overlaid
top - you'd think I'd learn by now, wouldn't you?)

Unimplemented syscall : 8
                         Untested (20092) [0x10171428]: syscall_kern.c line 696

	From unistd.h:
#define __NR_creat                8

	Not a clue.
	Cheers,
	- Bill

---------------------------------------------------------------------------
	"... being a Linux user is sort of like living in a house
inhabited by a large family of carpenters and architects. Every morning
when you wake up, the house is a little different. Maybe there is a new
turret, or some walls have moved. Or perhaps someone has temporarily
removed the floor under your bed." 
	-- Unix for Dummies, 2nd Edition
--------------------------------------------------------------------------
William Stearns (wst...@po...).  Mason, Buildkernel, named2hosts, 
and ipfwadm2ipchains are at:                http://www.pobox.com/~wstearns
LinuxMonth; articles for Linux Enthusiasts! http://www.linuxmonth.com
--------------------------------------------------------------------------

[User-mode-linux-devel] How many fhd's?

[William S. <wst...@po...>]

Good afternoon, Jeff,
	How many fhd's are supported?  It looks like MAX_DEV is set to 8
in the uml patch.
	Do they support partition tables so one might actually mount an
entire physical hard drive via fhd:

./linux fhd1=/dev/hdc

	where /dev/hdc is a partitioned ide drive?
	BTW - what do you think of the term Virtual Block Device as
opposed to Fake Hard Disk.  The former makes me think that I
could map vbda=/dev/hdc vbdb=/dev/ndb1 /vbdc=/dev/sr0 /vbdd=/dev/fd1 -
anything that is a block device, whereas fhd sounds like it can only be a
hard disk.
	Just thoughts...
	Cheers,
	- Bill

---------------------------------------------------------------------------
	Before you criticise a person, walk a mile in their shoes.  
That way, at least you're a mile away from them and you have their shoes.
--------------------------------------------------------------------------
William Stearns (wst...@po...).  Mason, Buildkernel, named2hosts, 
and ipfwadm2ipchains are at:                http://www.pobox.com/~wstearns
LinuxMonth; articles for Linux Enthusiasts! http://www.linuxmonth.com
--------------------------------------------------------------------------

[User-mode-linux-devel] 2.3.51 Networking *smile*

[William S. <wst...@po...>]

Good afternoon, Jeff and all,
	I'm working on networking for UML - I finally have it
functioning!  Thanks for the pointer, Jeff.
	I'll summarize what I've learned in a bit.  One note, though.  
After bringing the interface up, typing:
ifconfig umn down
	successfully returns to the command prompt, but promptly hangs uml
with a:
umn_read had error, errno = 9
	on the console and the idle thread at 99% on the host.

	It appears the uml patch hardcodes the address used on the host
side.  How would it be if, on the uml side, the user was expected to do:

ifconfig umn 172.20.0.244 pointopoint 172.20.0.253

	From which the uml-kernel could extract the pointopoint address
and execute

um_ifconfig sl0 172.20.0.253 pointopoint 172.20.0.244

	, rather than having to hardcode an address.  If two uml's were
running at the same time, then you could have 4 separate addresses, two
for each pointopoint link.

	I suppose Nirvanha would be to have sl0 on the host be able to
talk to multiple UML's on a simulated ethernet cable:
host OS sl0:	192.168.12.1/24
uml1 umn:	192.168.12.2/24, default route ->1
uml2 umn:	192.168.12.3/24, default route ->1
uml3 umn:	192.168.12.4/24, default route ->1
	Cheers,
	- Bill 'way too many damn suggestions' Stearns

---------------------------------------------------------------------------
	Any sufficiently advanced technology is indistinguishable from a
rigged demo. 
(Courtesy of Matthew Harrell <mha...@st...>)
--------------------------------------------------------------------------
William Stearns (wst...@po...).  Mason, Buildkernel, named2hosts, 
and ipfwadm2ipchains are at:                http://www.pobox.com/~wstearns
LinuxMonth; articles for Linux Enthusiasts! http://www.linuxmonth.com
--------------------------------------------------------------------------

[User-mode-linux-devel] Re: 2.3.51 panic - probable answer

[Jeff D. <jd...@ka...>]

> One more note; the same Kernel panic showed up when I was slowly
> filling up 24M of uml memory with executing bash over and over again.
> It seems that the hard drive space shrunk (perhaps as the memory used
> by uml grew?), and it seemed as if UML panic'ed right about the time,
> ahem, I ran out of disk space.

What I think is going on is that the hosting kernel only allocates memory (and 
disk space, since the memory is housed in an mmaped file) as the uml touches 
it.  When a disk or memory allocation fails, the uml suddenly finds that 
memory that it thought it had doesn't exist any more.

I've thought about this a bit more, and I think I'm going to consider it as 
equivalent to bad hardware, like a native kernel discovering that it has a 
bank of bad memory.  The native kernel will panic, and the lkml crowd will 
tell you to go away and replace your memory.

I'll see if I can come up with some better diagnostics, though.

				Jeff

[User-mode-linux-devel] Re: 2.3.51 panic

[Jeff D. <jd...@ka...>]

> Could you provide even the most meager instructions for what commands
> to manually type to allow me to ping the host OS from UML and to ping
> UML from the host OS and I'll take it from there?

This is somewhat hardwired at this point.  My network driver assumes that 
you're on 192.168.0.x right now.  Assuming this is true, this (inside) sets up 
the net:
	ifconfig umn 192.168.0.253 hw ether c0:a8:0:fd:0:0

The inside address is 192.168.0.253, so you can ping that from outside:

ping 192.168.0.253
PING 192.168.0.253 (192.168.0.253): 56 data bytes
64 bytes from 192.168.0.253: icmp_seq=0 ttl=255 time=124.7 ms
64 bytes from 192.168.0.253: icmp_seq=1 ttl=255 time=39.9 ms
64 bytes from 192.168.0.253: icmp_seq=2 ttl=255 time=49.9 ms

The outside address of the hosting kernel is 192.168.0.254 and you can ping 
that from inside:

ping 192.168.0.254
PING 192.168.0.254 (192.168.0.254): 56 data bytes
64 bytes from 192.168.0.254: icmp_seq=0 ttl=255 time=0.9 ms
64 bytes from 192.168.0.254: icmp_seq=1 ttl=255 time=99.8 ms
64 bytes from 192.168.0.254: icmp_seq=2 ttl=255 time=99.9 ms

Once this works, most other things should work.  telnet doesn't, though.  Web 
browsing does, and I've done it in both directions (lynx inside to hosting 
Apache and netscape outside to Apache inside).

				Jeff

[User-mode-linux-devel] Re: new syscall, nits, quirks, and blue sky.

[Jeff D. <jd...@ka...>]

> I take it this is the syscall for chroot?

Yup:

grep 61 include/asm-i386/unistd.h  
#define __NR_chroot              61

I'll toss that in.

> Hmmm - tough to preassign modes when devfs (in the host OS) creates
> them on the fly.  I'm sure there's some trickery in devfsd.conf that
> would cause this to happen, but it's not clear what.

For compatibility entries like these, I think the thing you're supposed to do 
is to make them, tar them up, and untar them into /dev at boot time.  That 
will let you keep permissions.

> 	I suspect that simply doing the above with two copies of the debian
> root_fs on fhd0 and fhd1 would _probably_ produce the same effect;

It would.  But that would be bad if they were mounted read-write.

> Specifically, when booting redhat, their [ OK ] and [ FAILED ] prompts
> change color

I tried to make this work.  I flipped on every color-related switch I could 
find in the xterm man page, and none of them did the trick.

> exec of "/sbin/rm" returned -2
> exec of "/sbin/rm" returned -2 
> exec of "/sbin/rm" returned -2

> /dev/xconsole: No such file or directory

At this point, if an error message doesn't seem to be harmful, I'm ignoring 
it.  And I've been ignoring those two for a while.  If anyone wants to figure 
out what's going on, feel free...

> 	What are the chances that I might someday be able to pass off host
> character or block devices to a uml incarnation?  For example: 
> ./ linux-2.3.51 ttyS0=/dev/ttyS0 

You can't do this yet.  It shouldn't be too hard, though.

> ./linux-2.3.51 fd0=/dev/fd0 

You can do that.  Did you try it?  Better to make it fhd1, though :-)  You can
also mount it inside the kernel because I compiled vfat into the binaries.  
CD-ROMs work, too and iso9660 is in there as well.

> all the uml threads label themelves with their job in the uml
> environment.  Nice touch!

I got sick of trying to figure out what all the threads were from the 
outside...

> Is there any thing I can do to help debug this?

First find the panicing thread, which you did (it's the one that goes into a 
tight loop):
> 9881 root      20   0  4124 4124  4120 R N  99.9 10.7   7:51 ./linux-2.3.51 [ls]

SIGUSR1 it : kill -USR1 9881

gdb linux and attach it.
get a backtrace.
get the faulting address from the segv frame.
if kern_segv_handler is on the stack, but not user_segv_handler, grab the 
value of its sc and "p *((struct sigcontext *)0xwhatever)".
then do "i sym 0xwhatever" and "i line *whatever" on sc->eip.

> Interestingly, a different pass shows Unknown HZ value! (19).

I don't know what's going on there.  Something seems to be calculating HZ and 
it seems not to be happy with what it sees.

> Could the routine that tries to allocate hard drive space for
> vm_file check to see if that allocation succeeded, perhaps?

It could.  But the kernel has to be able to deal with running out of memory.  
So I probably have some work to do.

> While hitting Ctrl-D's to exit out of those nested bash'es, uml hung
> and I got: Kernel panic: ptrace PTRACE_PEEKUSER returned 0, errno = 0

You hit some of the debugging code I put in to try to figure out your syscall 
0 problem.  It wasn't the code I hoped you'd hit, though :-(

				Jeff

[User-mode-linux-devel] 2.3.51 PTRACE_PEEKUSER panic

[William S. <wst...@po...>]

Good morning, Jeff,
	OK, one more and then I really have to go to bed.

	While hitting Ctrl-D's to exit out of those nested bash'es, uml
hung and I got:
Kernel panic: ptrace PTRACE_PEEKUSER returned 0, errno = 0
	overlaid over the "top" running on the uml console.
	Top on the host OS reports:
10621 root      20   0  1036 1036  1036 R    99.9  2.6  10:36 ./linux-2.3.51 [(tracing thread)]
	Sorry, not much else to offer.
	Cheers,
	- Bill

---------------------------------------------------------------------------
	Like the ad says, at 300 dpi you can tell she's wearing a
swimsuit. At 600 dpi you can tell it's wet. At 1200 dpi you can tell it's
painted on. I suppose at 2400 dpi you can tell if the paint is giving her
a rash. (So says Joshua R. Poulson) 
(Courtesy of Bob Taylor <brt...@qt...>)
--------------------------------------------------------------------------
William Stearns (wst...@po...).  Mason, Buildkernel, named2hosts, 
and ipfwadm2ipchains are at:                http://www.pobox.com/~wstearns
LinuxMonth; articles for Linux Enthusiasts! http://www.linuxmonth.com
--------------------------------------------------------------------------

[User-mode-linux-devel] Re: 2.3.51 panic - probable answer

[William S. <wst...@po...>]

Good morning, Jeff,

On Sun, 12 Mar 2000, William Stearns wrote:

> 	Host OS 2.3.51, rh6.2beta.  running uml-2.3.51 as root with debian
> root_fs.
> 	cd / ; ls -alR >/dev/null on first xterm.  Simultaneously
> trying to log into xterm2 and the uml console - the ls is chewing up a lot
> of cycles once the directory sectors have been loaded into ram.  On the
> console:
> 
> ...
> Debian GNU/Linux potato usermode ttys/0
> [User Mode Kernel v0.14]
> root password is "root"
> guest password is "guest"
> rootmode login:
> Password:
> Last login: Sun Mar 12 05:44:31 2000 on ttys/1
> Linux usermode 2.3.51-2um #5 Fri Mar 10 21:27:54 CST 2000 um unknown
> Kernel panic: Bogus address in segv
> ...
> 
> 	
> 	Try again, this time without "mem=20M" and "fhd1=rh6.2-root_fs" on
> the command line, i.e. just ./linux-2.3.51 .  Load up a couple of
> ls -alR >/dev/null 2>&1 &
> 's in memory (OK, 17 of them) and get another:
> Kernel panic: Bogus address in segv
> 	on the uml console, with 
>  9881 root      20   0  4124 4124  4120 R N  99.9 10.7   7:51 ./linux-2.3.51 [ls]
> 	showing up on the host OS' top.  Is there any thing I can do to
> help debug this?

	One more note; the same Kernel panic showed up when I was slowly
filling up 24M of uml memory with executing bash over and over again.  It
seems that the hard drive space shrunk (perhaps as the memory used by uml
grew?), and it seemed as if UML panic'ed right about the time, ahem, I ran
out of disk space.  OK, so it's pretty tough to even get the 200M
necessary to open up both root_fs's on a laptop.
	OK, redo the test with mem=16M - and I know I have 16M free on the
host drive that holds ~/uml .  start up multiple bash's on xterm1 and
watch the fun with the uml console.  I can almost fill up the available
16M, watching the host partition space shrink proportionately but never
get to 0, and as soon as I run the one bash that would fill it up, I get:
handle_mm_fault returned -1
Killing pid 144 because of a lack of memory
	on the console and:
...
$ bash
$ bash
$ bash
Bus error (core dumped)
$ 
	on xterm1.  Much better.

	It appears to be operator error.  Ahem - I need to remember to
have as much free hard drive space on the host partition as I want to
allocate for ram to UML.
	Could the routine that tries to allocate hard drive space for
vm_file check to see if that allocation succeeded, perhaps?
	Cheers,
	- Bill

[User-mode-linux-devel] 2.3.51 panic

[William S. <wst...@po...>]

Good morning, Jeff,
	Host OS 2.3.51, rh6.2beta.  running uml-2.3.51 as root with debian
root_fs.
	cd / ; ls -alR >/dev/null on first xterm.  Simultaneously
trying to log into xterm2 and the uml console - the ls is chewing up a lot
of cycles once the directory sectors have been loaded into ram.  On the
console:

...
Debian GNU/Linux potato usermode ttys/0
[User Mode Kernel v0.14]
root password is "root"
guest password is "guest"
rootmode login:
Password:
Last login: Sun Mar 12 05:44:31 2000 on ttys/1
Linux usermode 2.3.51-2um #5 Fri Mar 10 21:27:54 CST 2000 um unknown
Kernel panic: Bogus address in segv
...

	
	Try again, this time without "mem=20M" and "fhd1=rh6.2-root_fs" on
the command line, i.e. just ./linux-2.3.51 .  Load up a couple of
ls -alR >/dev/null 2>&1 &
's in memory (OK, 17 of them) and get another:
Kernel panic: Bogus address in segv
	on the uml console, with 
 9881 root      20   0  4124 4124  4120 R N  99.9 10.7   7:51 ./linux-2.3.51 [ls]
	showing up on the host OS' top.  Is there any thing I can do to
help debug this?


Cosmetic:
usermode:~# nice top
exec of "/usr/local/sbin/top" returned -2
exec of "/usr/local/bin/top" returned -2
exec of "/usr/sbin/top" returned -2
Unknown HZ value! (20) Assume 100.
	(then screen clears and top runs).  Interestingly, a different
pass shows Unknown HZ value! (19).  *smile*


Sincere request for help:
	Umm, I'm completely lost on how to set up the networking between
the host OS and uml.  Could you provide even the most meager instructions
for what commands to manually type to allow me to ping the host OS from
UML and to ping UML from the host OS and I'll take it from there?  I would
_really_ appreciate it.

	Thanks again!
	Cheers,
	- Bill

---------------------------------------------------------------------------
	Lavish spending can be disastrous.  Don't buy any lavishes for a 
while.
(Courtesy of Paul Jakma <pa...@cl...>)
--------------------------------------------------------------------------
William Stearns (wst...@po...).  Mason, Buildkernel, named2hosts, 
and ipfwadm2ipchains are at:                http://www.pobox.com/~wstearns
LinuxMonth; articles for Linux Enthusiasts! http://www.linuxmonth.com
--------------------------------------------------------------------------

[User-mode-linux-devel] new syscall, nits, quirks, and blue sky.

[William S. <wst...@po...>]

Good evening, Jeff,

On Sat, 11 Mar 2000, Jeff Dike wrote:

> > [wstearns@sparrow uml]$ touch vm_file 
> > [wstearns@sparrow uml]$ ./linux-2.3.49 devfs=nomount 
> > open: File exists 
> 
> The one thing that's important about vm_file is that it not exist beforehand.  
> That's the file that holds the kernel's physical memory.  It is opened (and 
> created) and then unlinked.

	Makes sense.

> What seems to be the problem is that your /dev/ttyp* files aren't usable by 
> anyone besides root.  Loosening up protections would help a lot.  Mine look 
> like this:
> 	crw-rw-rw-   1 root     tty        3,   0 May  5  1998 /dev/ttyp0

	Hmmm - tough to preassign modes when devfs (in the host OS)
creates them on the fly.  I'm sure there's some trickery in devfsd.conf
that would cause this to happen, but it's not clear what.  Oh well, use
root for the moment.

> Which reminds me that I'd better switch everything over to use /dev/pts 
> sometime...
> 
> Those diffs are innocuous.  Here's what they mean:
[snip]

	I have no problem believing that - thanks for the
explanation.


Unimplemented Syscall:
	And now to new business.  2.3.51-uml, debian root_fs-2.3.51,
started as root on a host 2.3.51.  I have the redhat root_fs loaded as
well (fhd1=rh6.2-root_fs on the command line, manually mounted inside
uml on /mnt/fhd1/; nifty trick, by the way!).
	I was trying to get a bash prompt completely inside the redhat
root partition, and thought that chroot would do the trick:
usermode:/etc/init.d# chroot /mnt/fhd1 /bin/bash
	uml hung on all three terminals.  Main console showed, slightly
mangled as I was running Midnight commander on it at the time:
Unimplemented syscall : 61
                          Untested (8987) [0x10171428]: syscall_kern.c line 696
	I take it this is the syscall for chroot?
	I suspect that simply doing the above with two copies of the
debian root_fs on fhd0 and fhd1 would _probably_ produce the same
effect; I don't _think_ this is some quirk of trying to switch between
debian and redhat.


Cosmetic:
	The terminals (console and both xterms) don't appear to be handle
some application requests.  Specifically, when booting redhat, their [ OK
] and [ FAILED ] prompts change color using, for example:
echo -en "\t\t\tWelcome to "
[ "$BOOTUP" != "serial" ] && echo -en "\\033[1;31m"
echo -en "Red Hat"
[ "$BOOTUP" != "serial" ] && echo -en "\\033[0;39m"
echo " Linux"
	to change the color of the words Red Hat.  Everything comes out
B&W.  Also, Midnight commander shows up (and even changes background color
via ncurses, but has incorrect line drawing characters on both the main
console and the xterms.  It also suffers from the same limitation that
SecureCRT on Windows has; the Function keys don't come through, so one
must use Esc-2 instead of F2.


Quirks:
...
Started device management daemon for /dev
Cleaning: /tmp /var/lock /var/runexec of "/sbin/rm" returned -2
exec of "/sbin/rm" returned -2
exec of "/sbin/rm" returned -2
...
	in booting.  Is this some search path quirk in Debian?  I've seen
this also in booting RH; it looks like that message may show up when an
application is searched for in each search directory.

...
INIT: Entering runlevel: 2
Starting system log daemon: syslogd  syslogd: /dev/xconsole: No such file
or directory
 klogd.
...
	I don't show /dev/xconsole on my system at all.  Is it needed?


Future:
	What are the chances that I might someday be able to pass off host
character or block devices to a uml incarnation?  For example:
./linux-2.3.51 ttyS0=/dev/ttyS0
./linux-2.3.51 fd0=/dev/fd0
	or perhaps better yet:
dd if=/dev/zero of=/home/wstearns/fakefloppy bs=1k count=1440
./linux-2.3.51 fd0=/home/wstearns/fakefloppy
	and have that host file accessed only when the uml mounts on its
own /dev/fd0?
	Granted, only one uml at a time could use a particular device in
the host, but...


2 points:
	Running top in the host OS is pretty cool - all the uml threads
label themelves with their job in the uml environment.  Nice touch!


	I've got to tell you, Jeff, it boggles my mind that one can do
this stuff at all!  Thanks again for all your work and patience with me.  
Let me know if I'm getting too nit-picky.
	Cheers,
	- Bill

---------------------------------------------------------------------------
	The best accelerator for a computer running MS Windows is FREE and
called gravity.
(Courtesy of Jos Hulzink <jo...@st...>)
--------------------------------------------------------------------------
William Stearns (wst...@po...).  Mason, Buildkernel, named2hosts, 
and ipfwadm2ipchains are at:                http://www.pobox.com/~wstearns
LinuxMonth; articles for Linux Enthusiasts! http://www.linuxmonth.com
--------------------------------------------------------------------------

[User-mode-linux-devel] Re: host os 2.3.51 and vm_file

[Jeff D. <jd...@ka...>]

> [wstearns@sparrow uml]$ touch vm_file 
> [wstearns@sparrow uml]$ ./linux-2.3.49 devfs=nomount 
> open: File exists 

The one thing that's important about vm_file is that it not exist beforehand.  
That's the file that holds the kernel's physical memory.  It is opened (and 
created) and then unlinked.

What seems to be the problem is that your /dev/ttyp* files aren't usable by 
anyone besides root.  Loosening up protections would help a lot.  Mine look 
like this:
	crw-rw-rw-   1 root     tty        3,   0 May  5  1998 /dev/ttyp0

Which reminds me that I'd better switch everything over to use /dev/pts 
sometime...

Those diffs are innocuous.  Here's what they mean:

> -fstat(4, {st_mode=S_IFREG|0775, st_size=16777217, ...}) = 0
> -mprotect(0x100c8000, 8192, PROT_READ|PROT_WRITE|PROT_EXEC) = 0
> +fstat(4, {st_mode=S_IFREG|0755, st_size=16777217, ...}) = 0
> +mprotect(0x100d6000, 8192, PROT_READ|PROT_WRITE|PROT_EXEC) = 0 

I don't know what the fstat is doing.  The mprotect is changing the 
permissions on the two pages containing init_task so that they are suitable 
for use as a stack.  The address is different because init_task is in the 
executable and it changed size from .49 to .51.

> -rt_sigaction(SIGUSR1, {0x100779dc, [], SA_RESTART}, {SIG_DFL}, 8) = 0
> -rt_sigaction(SIGIO, {0x100778b8, [], SA_RESTART}, {SIG_DFL}, 8) = 0
> -select(1, [], NULL, NULL, NULL)         = ? ERESTARTNOHAND (To be
> restarted) 
> ---- SIGUSR1 (User defined signal 1) --- 
> -kill(5671, SIGUSR1)                     = 0 
> -sigreturn()    = ? (mask now []) 
> -select(2, [0], NULL, NULL, NULL)        = ? ERESTARTNOHAND (To be restarted) 
> -+++ killed by SIGKILL +++ 
> +pipe([7, 8])                            = 0 
> +fn=0x411, child_stack=0x50001ff4, flags=0x7, args=0x1000000clone()
>                                = 6429 
> +rt_sigaction(SIGIO, {0x10084688, [], SA_RESTART}, {SIG_DFL}, 8) = 0
> +select(9, [7], NULL, NULL, NULL)        = 1 (in [7]) 
> +read(7, "\2\0\0\0\0\0\0\0\210\264\10\20\4\0\0\0\277\0\0\0000\n\0"..., 
> 		140) = 140 
> +select(9, [0 7], NULL, NULL, NULL 

I don't know what the SIGIO stuff is about here, but the rest of it is my 
redoing the mechanism by which other threads make requests of the input thread.

Its main job is to select on file descriptors that are important to the rest 
of the kernel and let it know when input has arrived.  Since it has the only 
user stack in the whole kernel and is the only thread that can call fork, I 
also gave it the job of firing off xterms for the virtual consoles.

I used to give it jobs by sending it a SIGUSR1 (hence the SIGUSR1 
disappearing) and have it figure out what's going on by having it read some 
global data.  Now it makes a pipe, listens on the read end, and any thread 
that needs something asks nicely on the write end.  This automatically 
serializes the requests and eliminates the need for the locking I was doing 
before.

The trace above is catching the input thread in the act of getting a request 
from the console to select on stdin (the read end of the pipe is fd 7, which 
gets input, and as a result of that input, the next select has fd 0 added to 
the read mask).
				Jeff

[User-mode-linux-devel] Re: host os 2.3.51 and vm_file - possible clue.

[William S. <wst...@po...>]

Good evening, Jeff,
	OK, so maybe _I_ don't have a clue, but... *smile*

On Sat, 11 Mar 2000, William Stearns wrote:

> 	Host: 2.3.51, shm, devpts, and devfs all mounted.
> 	Linux version 2.3.51-2um, rh62 root_fs (although I don't seem to
> get to the root_fs at all):
> ./linux-2.3.51 devfs=nomount
> ...
> Starting kswapd v1.6
> pty: 256 Unix98 ptys configured
> Initializing stdio console driver
> Initializing software serial port version 0
> Kernel panic: Out of pty's in getmaster (then hang...)
> 
> 	Same effect with 2.3.49, 2.3.49a, and 2.3.50a uml kernels, all on 
> 2.3.51 host.  Hmmm, something wrong with the host kernel?  2.3.51 has
> changes to shm - shared memory, perhaps?  shm is mounted on /var/shm .
> 	I've attached an strace of 2.3.49 trying to start up.  It gives
> the above messages.  The trace seems to be concerned with a file called
> vm_file, so lets try a game:
> [wstearns@sparrow uml]$ touch vm_file
> [wstearns@sparrow uml]$ ./linux-2.3.49 devfs=nomount
> open: File exists
> 	Hmmm - does vm_file have some problem opening on 2.3.51?
> 	I'll play some more.

	OK, more clues.  When I start 2.3.51 as root, it starts:
...
Starting kswapd v1.6
pty: 256 Unix98 ptys configured
Initializing stdio console driver
Initializing software serial port version 0
serial line 0 assigned pty /dev/ptyp0
ssl receive thread is pid 6338
devfs: v0.93 (20000306) Richard Gooch (rg...@at...)
...
	and comes up just fine (well, other details, but I'm looking into
those).  Can anyone else reproduce this?  I'd be happy to provide a trace
of the successful (run as root on host 2.3.51) pass on demand, but in the
interest of brevity, here's the diff of the ends of the traces.  Keep in
mind the first ("-") is 2.3.49, and the second ("+") is 2.3.51.

 lseek(4, 16777216, SEEK_SET)            = 16777216
 write(4, "\0", 1)                       = 1
 old_mmap(0x50000000, 16777216, PROT_READ|PROT_WRITE,MAP_SHARED|MAP_FIXED, 4, 0) = 0x50000000
-fstat(4, {st_mode=S_IFREG|0775, st_size=16777217, ...}) = 0
-mprotect(0x100c8000, 8192, PROT_READ|PROT_WRITE|PROT_EXEC) = 0
+fstat(4, {st_mode=S_IFREG|0755, st_size=16777217, ...}) = 0
+mprotect(0x100d6000, 8192, PROT_READ|PROT_WRITE|PROT_EXEC) = 0
 mprotect(0x50001000, 4096, PROT_READ|PROT_WRITE|PROT_EXEC) = 0
-fn=0x411, child_stack=0x50001ff4, flags=0x7, args=0x1000000clone()                                 = 5671
-rt_sigaction(SIGUSR1, {0x100779dc, [], SA_RESTART}, {SIG_DFL}, 8) = 0
-rt_sigaction(SIGIO, {0x100778b8, [], SA_RESTART}, {SIG_DFL}, 8) = 0
-select(1, [], NULL, NULL, NULL)         = ? ERESTARTNOHAND (To be restarted)
---- SIGUSR1 (User defined signal 1) ---
-kill(5671, SIGUSR1)                     = 0
-sigreturn()                             = ? (mask now [])
-select(2, [0], NULL, NULL, NULL)        = ? ERESTARTNOHAND (To be
restarted)
-+++ killed by SIGKILL +++
+pipe([7, 8])                            = 0
+fn=0x411, child_stack=0x50001ff4, flags=0x7, args=0x1000000clone()                                 = 6429
+rt_sigaction(SIGIO, {0x10084688, [], SA_RESTART}, {SIG_DFL}, 8) = 0
+select(9, [7], NULL, NULL, NULL)        = 1 (in [7])
+read(7, "\2\0\0\0\0\0\0\0\210\264\10\20\4\0\0\0\277\0\0\0000\n\0"...,
140) = 140
+select(9, [0 7], NULL, NULL, NULL
\ No newline at end of file

	Possible relevant differences: the mode on the early fstat,
additional pipe in 2.3.51, SIGUSR1 handler in 2.3.49, perhaps?
	Cheers,
	- Bill

---------------------------------------------------------------------------
	Weinberg's Law: If builders built buildings the way programmers
wrote programs, then the first woodpecker that came along would destroy
civilization. 
(Courtesy of David E. Vandewalle, van...@pr...)
--------------------------------------------------------------------------
William Stearns (wst...@po...).  Mason, Buildkernel, named2hosts, 
and ipfwadm2ipchains are at:                http://www.pobox.com/~wstearns
LinuxMonth; articles for Linux Enthusiasts! http://www.linuxmonth.com
--------------------------------------------------------------------------

[User-mode-linux-devel] host os 2.3.51 and vm_file

[William S. <wst...@po...>]

Good evening, Jeff,
	Saturdays are much easier for finding time for the _fun_
stuff.  *smile*

	Host: 2.3.51, shm, devpts, and devfs all mounted.
	Linux version 2.3.51-2um, rh62 root_fs (although I don't seem to
get to the root_fs at all):
./linux-2.3.51 devfs=nomount
...
Starting kswapd v1.6
pty: 256 Unix98 ptys configured
Initializing stdio console driver
Initializing software serial port version 0
Kernel panic: Out of pty's in getmaster (then hang...)

	Same effect with 2.3.49, 2.3.49a, and 2.3.50a uml kernels, all on 
2.3.51 host.  Hmmm, something wrong with the host kernel?  2.3.51 has
changes to shm - shared memory, perhaps?  shm is mounted on /var/shm .
	I've attached an strace of 2.3.49 trying to start up.  It gives
the above messages.  The trace seems to be concerned with a file called
vm_file, so lets try a game:
[wstearns@sparrow uml]$ touch vm_file
[wstearns@sparrow uml]$ ./linux-2.3.49 devfs=nomount
open: File exists
	Hmmm - does vm_file have some problem opening on 2.3.51?
	I'll play some more.
	Cheers,
	- Bill

---------------------------------------------------------------------------
	"She worked with a subdued intensity... She once told me that the
only way to know when you have done something truly great is when your
spine tingles."
	- on Alice Kober, cryptanalist, in The Code Book, Simon Singh.
--------------------------------------------------------------------------
William Stearns (wst...@po...).  Mason, Buildkernel, named2hosts, 
and ipfwadm2ipchains are at:                http://www.pobox.com/~wstearns
LinuxMonth; articles for Linux Enthusiasts! http://www.linuxmonth.com
--------------------------------------------------------------------------

[User-mode-linux-devel] New cvs checkins

[Jeff D. <jd...@ka...>]

I changed the mechanism that other threads use to make requests of the input 
thread.  Instead of signalling it, they now send messages over a pipe that it 
sets up.

The console driver now kills off xterms when the open count goes to zero 
rather than having the machine shutdown process kill them.

I added a couple more build tweaks from Peter Moulder.

HZ is now 20, rather than 10.  This lets me remove include/net/tcp.h from the 
patch since Lars apparently prevailed on Alexey to support HZ==20.  It also 
seems to have doubled the BogoMips (which I don't understand - it should have 
cut it in half) and who can argue with that?

				Jeff

[User-mode-linux-devel] user-mode port 2.3.50

[Jeff D. <jd...@ka...>]

I've updated the user-mode kernel to 2.3.50.
Other developments include:
	eliminated the unexpected SIGIOs that the input thread was getting
	a much improved config and build process, thanks to Peter Moulder
	a readonly option on the block device driver, ditto
	fixed the network driver, which I broke a couple of releases ago during the 
responsiveness fixes

It's also checked into cvs ( http://sourceforge.net/cvs/?group_id=429 ).

The patch and kernel have been updated and are available from 
http://sourceforge.net/project/filelist.php?group_id=429

				Jeff

[User-mode-linux-devel] Invalid session number?

[Jeff D. <jd...@ka...>]

While testing the new readonly device support, I ran linux like this:
	linux fhd0=../../root_fs_deb fhd1=swap fhd2r=/dev/hdc fhd3r=/dev/fd0

where hdc is my CD-ROM. When I mounted it in the user-mode kernel, I got:

usermode:/# mount /dev/disk/2 /mnt
mount: block device /dev/disk/2 is write-protected, mounting read-only
Invalid session number or type of track
Invalid session number

Aside from that error message, it seems fine.  Everything is there.  But I 
don't get that message when I mount it in the physical machine.

Does anyone know what might be going on?

BTW, this is with iso9960 and vfat in the kernel, and they both seem to work 
fine.

				Jeff

[User-mode-linux-devel] The latest code is checked in to cvs

[Jeff D. <jd...@ka...>]

It's still at 2.3.49, and will be until that Linus slacker releases 2.3.50...

The patch and kernel on the download page are current as well.

I think the SIGIO problem is gone.
A number of system calls are now hooked up.
mkrootfs now turns on two virtual consoles again now that the xterm memory 
corruption problem seems to be fixed.
There's some new code in the system call path to try to track down William 
Stearns syscall 0 problem.

				Jeff

[User-mode-linux-devel] none already mounted (was Re: uml 2.3.49 results.

[Peter M. <re...@ne...>]

> > mount: none already mounted or /dev/pts busy mount: according to mtab,
> > devpts is already mounted on /dev/pts ... (any relevance?) 

I think I fixed this by changing `none' in /etc/fstab to `proc' and
`devpts' respectively.

pjm.

[User-mode-linux-devel] Re: more uml 2.3.49 results and a first pass rpm of the root filesystem.

[Jeff D. <jd...@ka...>]

wst...@po... said:
> Unimplemented syscall : 197
>                            Untested (2216) [0x10154c48]:
> syscall_kern.c line 672 

That's fstat64.  I was wondering when I was going to start seeing the 64-bit 
system calls.

I've got a new kernel brewing with the SIGIO fix, some code to try to trap the 
syscall 0 thing, and a few more syscalls hooked up.  I'll add fstat64 and send 
it to sourceforge, and let you know when it's there.

> Does the root_fs need to be contiguous, or can the uml kernel open a
> root_fs that's not?

If you did dd of=filename rather than dd of=devname, then you got a normal, 
not necessarily contiguous file.  The kernel doesn't care.

> I'd suggest that.  Have a uml-rhroot-6.2, uml-rhroot-6.0,
> uml-debroot-2.0, etc. rpms for the root and uml-kernel-2.3.49,
> uml-kernel-2.3.46, etc.

I was originally against this, on the grounds that you don't change 
environments all that often, and when you do, you might as well grab the 
appropriate package, along with the latest kernel, since it doesn't add much 
in terms of download time.

The idea has grown on me, though.  If we get a good variety of enviromnents, 
then people might switch between them more often, and this packaging makes it 
clear that it's possible.

I would prefer to ship them as compressed tar files, though.  People wanting 
to try out a RH environment aren't necessarily going to have rpm available.

				Jeff

[User-mode-linux-devel] more uml 2.3.49 results and a first pass rpm of the root filesystem.

[William S. <wst...@po...>]

Good morning, Jeff,

On Sun, 5 Mar 2000, Jeff Dike wrote:

> > 	This one's not too hard to trigger - let me know if I can supply
> > additional info or a shell account on this laptop. 
> 
> I can't reproduce it at all.  I've done your things, I've done some other 
> strange things, I've done a bunch of them running at the same time, and all I 
> get is pages of numbers and no crashes.

	I give it a 60% chance that it's some odd quirk in my host
environment.  This is a laptop with pcmcia, apm enabled, probably a number
of other quirks a desktop machine wouldn't have.
	Supporting that theory is the fact that I find that a shell app
I've written tends to crash from time to time; I think it's getting some
kind of signal it doesn't expect.
	We keep looking.

stock 2.3.49 (Linux version 2.3.49-1um (jd...@cc...)), rh6.2
root (although we haven't gotten to that yet:

VFS: Mounted root (ext2 filesystem) readonly.
Mounted devfs on /dev
INIT: version 2.78 booting
Unimplemented syscall : 197
                           Untested (2216) [0x10154c48]: syscall_kern.c line 672

	and again, with:
INIT: version 2.78 booting
Unimplemented syscall : 197
                           Untested (2403) [0x10154c48]: syscall_kern.c line 672

	, add "devfs=nomount" to the command line as I forgot to add
devfsd to the rootfs and:
INIT: version 2.78 booting
Unimplemented syscall : 197
                           Untested (2440) [0x10154c48]: syscall_kern.c line 672

	Sorry, can't get it to boot.  Hmmm.... wait a second!  In making
the root_fs in the rpm, I used dd to get a contiguous block of
sectors.  Once packaged and installed, that is almost certainly no longer
the case.
	Does the root_fs need to be contiguous, or can the uml kernel
open a root_fs that's not?

> > usermode login: thread got SIGIO 
> 
> I think I fixed this one, though.  I didn't read the fcntl man page carefully 
> enough for my last fix of this problem.

	Thanks...

> > Off to try to get an RH6.2beta root_fs ready.  Do you have any
> > interest in posting it to uml.sourceforge? 
> 
> Yes, I do.  I just haven't run into any RH media later than 6.0 recently.

	I pulled down the rpm's and stuck them in a directory and built
the rpm.  I've attached the src rpm; if anyone would like to try it, put
it in /usr/src/redhat/SRPMS, put the necessary rpms (list attached
above, versions will be different if you're using a distribution other 
than rh6.2beta) in /mnt/flipper/usr/src/uml-rhrpms/ and:

cd /usr/src/redhat/SRPMS
rpm --rebuild uml-rhroot-6.2-0.src.rpm

	For those that want to see the guts, the spec file will be left in
/usr/src/redhat/SPECS/ and the various configuration files will be in
/usr/src/redhat/SOURCES/ .
	This is _not_ a final rpm, just being made available to those that
want to play.

> > I'd like to sincerely
> > suggest making two rpm's, one for the uml kernel, one for the root_fs
> > itself, so people can mix and match.
> 
> How's that different from what I'm doing now, except that the packages include 
> the kernel?  Are you suggesting a package without a kernel?

	I'd suggest that.  Have a uml-rhroot-6.2, uml-rhroot-6.0,
uml-debroot-2.0, etc. rpms for the root and uml-kernel-2.3.49,
uml-kernel-2.3.46, etc.  The user can mix and match by copying their
root_fs of choice and their kernel of choice from /var/lib/uml to a 
working directory.
	How does that sound?
	Cheers,
	- Bill

---------------------------------------------------------------------------
	I like cats too, lets exchange recipies.
(Courtesy of John Michael Clemens <cl...@rp...>)
--------------------------------------------------------------------------
William Stearns (wst...@po...).  Mason, Buildkernel, named2hosts, 
and ipfwadm2ipchains are at: http://www.pobox.com/~wstearns/
--------------------------------------------------------------------------

[User-mode-linux-devel] Re: uml 2.3.49 results.

[Jeff D. <jd...@ka...>]

> usermode login: thread got SIGIO
>                 ^^^^^^^^^^^^^^^^
> 	The above showed up a soon as I hit "r" to log in as root. 

I could have swore I fixed that...  That's been bothering me for a while, and 
I don't have any idea why it's happening.

> /dev/disk/0 was not cleanly unmounted, check forced. Kernel panic:
> Bogus address in segv 

This is another known one.  I don't see it often enough to have an idea how to 
start tracking it down.

> mount: none already mounted or /dev/pts busy mount: according to mtab,
> devpts is already mounted on /dev/pts ... (any relevance?) 

I don't think so.  It seems harmless.

> usermode:/# ls -alR | wc -l 
> ls: ./proc/2/exe: No such file or directory 
> ls: ./proc/3/exe: No such file or directory 
> ls: ./proc/4/exe: No such file or directory 
> Unimplemented syscall : 0
>                       Untested (1521) [0x10154c48]: syscall_kern.c line 672 

This is just strange.  I've never seen anything like this before.

> 	This one's not too hard to trigger - let me know if I can supply
> additional info or a shell account on this laptop. 

Good, because this is something I definitely need to fix.

> 	I still get strange warnings from inetd when the xterms are
> re-enabled, but the xterms start just fine: 
> Starting internet superserver: inetd. 
> error stat'ing /dev/ttydirs/xxx/xxxxxxxxxxxxp1: No such file or directory
> error stat'ing /dev/ttydirs/xxx/xxxxxxxxxxxxp2: No such file or directory 

They're not coming from inetd.  They're coming from somewhere else.  inetd 
just happens to look guilty by association.


				Jeff